Professional Documents
Culture Documents
PA T R I K N . J U S L I N
UPPSALA UNIVERSITY
He put the bow to his instrument . . . and then, the first notes, bold and fiery,
sang through the hall. At once the spell began to work. Was this really the
music of a violin? What grandeur in these slurred notes, what absolute purity!
There came roulades of double-stop harmonic notes, and a long run across four
octaves, played staccato in a single stroke of the bow . . . Then came a noble,
moving theme, which sounded as though a human voice was singing . . . After
the seemingly endless applause had subsided, Paganini began to play the second
movement. It was an adagio, and showed the virtuoso from quite a different
angle. There were none of the devilish tricks that had stunned the audience
during the first movement. A sublime, angelic song of great noblesse and sim-
plicity touched the hearts of the listeners . . . The notes followed one another as
sempre :
274 Psychology of Music 31(3)
though growing out of the instrument, and it seemed incredible . . . that this
wooden object was not an integral part of the man who played it, a part of his
very soul . . . The audience sat as though paralysed until the rhythm of a grace-
ful rondo changed their mood . . . an infinitely tender pizzicato accompanied the
melody, and it finally soared away into a happy dance tune. (Farga, 1969:
171–2)
the fact that musical expression involves tacit knowledge that is difficult to
convey from teachers to students. Research on expression could help to ren-
der the tacit knowledge explicit. Rather than surrender to individual differ-
ences in expressive ability in the name of musical talent, we could address
them through theoretically informed teaching. Thus, a proper and scientifi-
cally grounded understanding of the mechanisms that underlie musical
expression has scientific as well as social and educational implications
(Sloboda, 2000).
In this article, I consider the role of the performer in musical expression. I
want to convince the reader that performance expression is a problem
amenable to empirical investigation, and that psychological theory is critical
to an understanding of this problem. Looking back, psychologists have tend-
ed to approach performance expression mainly by measuring various
acoustic variables of music performances (e.g. Gabrielsson, 1999, 2003, this
issue). Often, a purely descriptive approach has been taken, in accordance
with the tradition first established by Seashore (1938) and his co-workers. As
a consequence, the notion of expression has been poorly conceptualized.3
Matters of expression have largely been reduced to tables or graphs of
acoustic data, whereas the question of what these data actually tell us about
the origins of musical expression has somehow been lost. But perhaps
psychology is uniquely suited to investigating expression? After all, psycholo-
gy was right from the beginning studying the kinds of psychophysical
relationships that seem to underlie musical expression (Leahey, 1987). Yet,
psychology has arguably not fulfilled its potential role in helping to provide a
theoretical organization of the field. In my view, a psychological approach to
performance expression should consider how this phenomenon reflects basic
human abilities and characteristics that are not necessarily unique to the
musical domain. Recent research reviewed in this article will provide some
support for this view.
The remainder of this article is structured as follows: first, I provide
working definitions of interpretation, expression and communication, and
consider problems in performance research that have prevented cumulative
progress. Second, I outline a psychological approach to musical expression
that distinguishes five aspects of expression. Unfortunately there are few
attempts to integrate these different aspects; and so, I outline a meta theory of
music performance, which is an attempt to integrate different aspects of
expression. Not very surprisingly, empirical findings derived from this model
suggest that emotion is a major aspect of expression. Therefore I briefly
review findings about emotion in music performance. Finally, I consider
various implications of a psychological approach for performance research
and music education, and also acknowledge some limitations of this
approach.
276 Psychology of Music 31(3)
Working definitions
Interpretation, expression and communication are complex concepts, and no
matter what I say about them here, I am sure that this article will not be the
final word on these topics. However, for the practical purposes of this article,
we need to have at least preliminary working definitions of each term. The
term interpretation typically refers to the individualistic shaping of a piece
according to the musical ideas of the performer (Palmer, 1997: 119). This
might involve an intention to express something (e.g. an idea, an emotion)
beyond the musical structure (Gabrielsson, 1999), but such expression is usu-
ally accomplished through the ways in which the structure is articulated (in
how it is played). The process of interpretation is still little investigated (but
see Hallam, 1995; Persson, 1993), although it seems to be influenced by both
‘internal’ (e.g. emotions, wanting to express something personal) and ‘exter-
nal’ factors (e.g. musical style, the structure of the piece, the composer’s
intentions) (Lindström et al., in press; Persson, 2001: 278–81).
Expression, in my view, refers to a set of perceptual qualities that reflect
psychophysical relationships between ‘objective’ properties of the music, and
‘subjective’ (or, rather, objective but partly person-dependent) impressions of
the listener. Expression does not reside solely in the acoustic properties of the
music (different listeners may perceive the expression differently), nor does it
reside solely in the mind of the listener (different listeners usually agree about
the general nature of the expression in a performance). Expression depends
on both of these factors, in ways that, although complex, can be modelled in a
systematic fashion (Juslin, 2000). Our perception of expressive music per-
formances reminds us, somehow, of the ways humans express their states of
mind in real life (why would we otherwise use the term expression in the first
place?). Indeed, we have an ‘anthropomorphic’ tendency to perceive expres-
sive form even in inanimate objects (e.g. Eibl-Eibesfeldt, 1989). Occasionally
the perception of an expressive performance will also evoke an emotion in the
listener (he or she is ‘moved’), or even an aesthetic response (a more complex
and profound response that involves emotional, cognitive and social factors;
Koneĉni, 1979), but such a response is not required for a listener to hear the
music as expressive (Davies, 1994). The emotion induced in the listener may
be the same as that expressed (through emotional contagion; Juslin, 2001),
or it may be a complementary response (Juslin and Zentner, 2002). As Clynes
(1977: 60) notes, when we hear a convincing expression, this is perceived as
‘sincerity’; we therefore tend to feel sympathy for the performer of the music.
If the general notion of expression is accepted, the question is: what does
music express. A survey of the literature suggests a large number of different
ideas about what music may express. Music has been regarded as expressive
of emotion, physical aspects (motion, force), tension and release, personality
characteristics, beauty, events, objects, musical conventions, religious belief
and social conditions (Gabrielsson and Juslin, 2003). A generous view on
Juslin: Five facets of musical expression 277
expressivity would hold that there is truth to all of these ideas as long as there
are individual listeners who claim that they find music expressive of these
qualities. One problem with this generous approach is that it tends to
embrace any arbitrary association that one could have to a music performance.
This would be simply a form of conditioning that does not relate to the per-
formance as such; it says more about the listener than it says about the
music. It seems more fruitful to look for expressive aspects that relate to the
nature of the music, at least if we want to explain performance expression.
A more restrictive approach to expression that is common in research
holds that music is expressive of a certain quality only to the extent that there
is some minimum level of agreement among the listeners (presumably
because there is something in the actual performance that gives rise to simi-
lar listener percepts). Thus, for example, we might agree that the slow move-
ment of Beethoven’s Eroica symphony is expressive of, say, sadness (although
the expression could, of course, change rapidly during the piece). In such
cases of intersubjective agreement, we may be inclined to say that ‘this piece
has a sad expression’. But note that expression (as defined here) does not
require an expressive intention.
The concept of communication (of emotion, for instance), in contrast, goes
further: accurate communication, I believe, requires that there is both a per-
former’s intention to express a specific concept and recognition of this concept
by a listener. Perhaps, it may seem strange to talk about communication
accuracy in the context of music. Still, most performers are probably – or
should be – worried about whether their musical interpretation is actually
perceived by listeners the way they intended it. (What is the purpose of a spe-
cific interpretation if every listener fails to perceive it?) The performer may, for
instance, wish to highlight an emotional character that is latent in the com-
position. The extent to which performer and listener agree about the
emotional expression of the performance could pragmatically be seen as a
measure of the accuracy of the communication. (Precisely which perceptual
qualities I think are involved in performance expression is discussed later in
the description of a psychological approach.)
Context-related Acoustics
Sound technology
Listening context (e.g. recording, concert)
Other individuals present
Visual performance conditions
Larger cultural and historic setting
Whether the performance is formally evaluated
Juslin: Five facets of musical expression 279
performances from expert musicians (however, see Repp, 1998). This prob-
lem can be partly overcome through large investments of time and money,
and also by trying to conduct research more on the performers’ own terms.
Second, even if we do manage to obtain sufficiently large samples of perform-
ances, it is a difficult and time-consuming task to analyse all relevant
acoustic variables (Palmer, 1997). There are currently some promising
attempts to develop computer algorithms for automatic extraction of such vari-
ables (e.g. Friberg et al., 2002), which might help researchers to analyse larg-
er samples of performances quickly. Thus this problem may eventually be
solved. The most serious problem for studies of performance is perhaps not to
obtain sufficient amounts of data, but rather to find ways of interpreting the
wealth of data in a meaningful way (Gabrielsson and Juslin, 1996).
It is here that psychological theory can make a crucial contribution to how
we conceive of expression. I think that the greatest impediment to progress in
explaining expression has been the common tendency to regard expression as
a single entity, a homogeneous natural category. Given the controversy that
has surrounded the concept of expression throughout history (Davies,
1994), it is perhaps understandable that many researchers have tended to
leave the concept undefined or simply ‘defined’ it in terms of ‘deviations from
the score’. But it is troubling to note how many studies have treated expres-
sion as a mysterious quality of which there is simply ‘more’ or ‘less’, without
specifying what is meant by the term expression (the expression is simply
‘appropriate’, ‘exaggerated’, or ‘lacking’). There is no serious consideration of
what is expressed, or how it is expressive, which implies that there is only one
way of performing expressively (by ‘appropriate expressive deviations’). At
times, it seems that expression is simply equated with everything that might
be good about a performance, but what is good is, of course, not specified.
What is at stake is not the usefulness of the generic word ‘expression’ as such,
it is rather whether this is the most fruitful level at which to investigate and
teach expression (‘Put some expression into it’). I reject on empirical grounds
the idea that music can be expressive ‘in general’ without being expressive in
some particular way, because any music performance has certain acoustic
features that renders it different in expression from a performance with differ-
ent acoustic features. (Two performances with the same rated level of
‘expressivity’ or ‘emotionality’ may exhibit quite different acoustic character-
istics, and may be perceived quite differently with regard to how they are
expressive.)
Even more important, the expressive variations that are typically found in
a human music performance are not all of the same kind; they do not all
share the same origins, and they should therefore not all be taught in the
same manner. In my estimation, the ‘single-factor approach’ to expression
has contributed to the fragmented state of affairs in studies of performance. It
has not encouraged integrative attempts even among researchers who have
recognized that expression is multidimensional. I argue that a satisfactory
280 Psychology of Music 31(3)
Note. The solid line shows the first performance, the dotted line the repetition.
(Reproduced with kind permission of the Royal Swedish Academy of Music)
FIGURE 1 Deviations in timing from mechanical performance in a performance of Mozart’s
Piano Sonata in A Major (K 331) from a classic study by Gabrielsson (1987).
Juslin: Five facets of musical expression 281
GENERATIVE RULES ( G )
One function of performance expression may be to convey the musical struc-
ture to listeners as clearly as possible. This forms the basis of the so-called
generative approach (Clarke, 1988). In this line of research, expression is
regarded as rule-based transformations of nominal score values that origi-
nate in the performer’s cognitive representation of the hierarchical structure.
By means of variations in such acoustic variables as timing, dynamics and
articulation, a performer is able to clarify group boundaries (Gabrielsson,
1987), metrical accents (Sloboda, 1983) and harmonic structure (Palmer,
1996). One of the most robust findings is that tempo variations (rubato) have
a strong tendency to be determined by the phrase structure; phrase endings
are usually marked with decreases in tempo, and the amount of slowing
reflects the depth of embedding in the hierarchical structure (Todd, 1985; see
also Figure 1). This is the most well-studied aspect of expression, and to the
extent that a performer aims to clarify the structure, we may explain part of
the variance in the acoustic features. However, the absence of a generally
accepted system for structural analysis still makes prediction difficult.
282 Psychology of Music 31(3)
EMOTIONAL EXPRESSION ( E )
A second function of performance expression might be to render the perform-
ance with a particular emotional expression. As noted by Shaffer (1992), ‘a
performer can be faithful to the structure and at the same time have the free-
dom to shape its moods’ (p. 265). A series of studies in the 1990s indicated
that professional performers are able to communicate discrete emotions to lis-
teners. To accomplish this goal, the performers use a large number of
acoustic variables in the performance. For example, if a performer would like
to express tenderness in a performance, he or she may use slow tempo, low
sound level, legato articulation, ‘soft’ timbre, slow tone attacks, regular tim-
ing, reduced contrasts between ‘long’ and ‘short’ notes, and an intense vibra-
to (for a review, see Juslin, 2001). This component of expression is actually
the last to receive attention in performance research, but it could well be one
of the most crucial as far as music performers and listeners are concerned
(see Lindström et al., in press; Persson, 2001). One reason for the relative
neglect of this component may be that it not always operates on the explicit
level that is required in order to study it empirically. (The emotion component
is discussed in more detail in a later section of this article.)
RANDOM VARIABILITY ( R )
A third aspect of performance expression that has mostly been studied in
domains other than music is random variability. It appears to be generally
agreed that human music performance is not controlled by a completely
deterministic motor-system (e.g. Repp, 1997a; Yamada, 1998). Performance
expression always contains some random fluctuations, although they may be
quite small. How is this feature relevant in understanding expression, espe-
cially since practice usually aims at minimizing error? From an aesthetical
point of view, random variations contribute to the ‘living’ character of music
– that slight unpredictability that makes each performance absolutely
unique. Research on human limitations in perceptual-motor skills, mainly
studies of isochronous interval finger tapping, have revealed that the random
variations have certain characteristics: (1) the magnitude of random fluctua-
tions increases with inter-onset-interval duration, so that longer intervals
tend to yield larger deviations (Wing and Kristofferson, 1973; see also Repp,
1997a); (2) there is a negative first-order dependency with regard to inter-
onset-interval durations; that is, a performed interval that is shorter than the
mean is usually followed by one that is longer than the mean, yielding zig-zag
Juslin: Five facets of musical expression 283
MOTION PRINCIPLES ( M )
A fourth aspect of performance expression is motion. There is a common
assumption that music and motion are closely related to one another.
However, this hypothesis needs to be made more specific to be useful. Shove
and Repp (1995) argue that ‘without proper constraint the idea that change
in music induces an experience of motion has little explanatory power’ (p. 58).
One way to constrain the hypothesis is to limit it to a specific kind of motion
called biological motion (e.g. Johansson, 1973). This refers to the dynamic
patterns of movement that are characteristic of humans. It should be noted
that such patterns in music performance may be of two kinds. First, it can be
assumed that performers intentionally (though not necessarily consciously)
try to re-create such patterns. Shove and Repp propose that an aesthetically
pleasing performance is ‘one whose expressive microstructure satisfies basic
constraints of biological motion’ (p. 78). One example is the shaping of final
ritardandi. Friberg and Sundberg (1999) showed that final ritardandi of
music performances follow a mathematical function similar to that of run-
ners’ decelerations. (This was also the ritardando function preferred by listen-
ers in a listening test.) A second kind of biological motion is non-intentional
patterns of variability that reflect anatomical constraints of the body in con-
nection with motor requirements of specific musical instruments (e.g. Penel
and Drake, 1999).
STYLISTIC UNEXPECTEDNESS ( S )
An additional way in which a performance may be expressive involves the
fact that musical emotions often occur when musical expectations are violat-
ed in some way. Thus, Meyer (1956) suggested that expressive variations in a
music performance may serve an aesthetic function by ‘delaying an expected
resolution’, or otherwise ‘creating psychological tension’ (p. 206). This could
happen when a performer deviates from stylistic expectations with regard to
performance conventions for a certain part of the structure. For example, the
performer might be expected to clarify the structure in a certain manner (as
described by generative rules), but instead he or she does something com-
pletely different. This momentarily creates psychological tension that is
resolved when the performer resumes ‘expected playing’. Something of the
284 Psychology of Music 31(3)
This component is probably the least researched so far, but it may be criti-
cal to developing a truly original interpretation. For examples of music per-
formances by experts that feature various bold and unexpected patterns of
expression, see, for example, Repp (1997b). No attempt has yet been made to
model this component in terms of expressive rules, perhaps because of the
paradox inherent in developing rules for how to break other rules, and the
difficulty in deciding precisely on what basis such rule breaking is done (not
all rule breaking is musically satisfying).
Component
Characteristic G E R M S
Origin of pattern Generative Emotion-specific Internal timekeeper Biological motion; Deviations from
transformations of patterns of acoustic and motor delay distinct patterns of expected performance
the musical structure cues deriving from variance reflecting movement typical conventions
vocal expression human limitations of human beings
Nature of pattern Local expressive Mainly overall levels Semi-random patterns Dynamic, non- Local; not predictable
features related to of multiple uncertain, 1/f noise and white compensatory from the structure
the structural partly redundant cues noise; very small in patterns; smooth
interpretation that are compensatory magnitude, irregular and global
Salient brain regions Left hemisphere Right hemisphere Lateral and medial Left hemisphere Anterior cingulate
(adjacent to Broca’s (the basal ganglia) parts of the (adjacent to the cortex
area) cerebellum, plus the superior temporal
motor cortex sulcus)
Perceptual effects Clarifies structure; Expresses emotions Generates a ‘living’ Yields expressive Heightens tension
affects the inherent and moods (mainly in and natural quality form that is similar and unpredictability
expression of a piece broad categories of to human gestures
emotion)
Knowledge dependence Medium Low None Low High
Aesthetic contribution Beauty, order, Recognition, arousal, Unevenness, novelty Balance, unity, Novelty, arousal
coherence personal expression recognition
Under voluntary control Yes, mostly Yes No Yes, partly Yes
Juslin: Five facets of musical expression 287
A PRELIMINARY IMPLEMENTATION
Can this kind of psychological theory be turned into something useful in
empirical terms? In a recent study, we made the first attempt to implement a
component approach in a computational model, the GERM model, that simu-
lates different aspects of expression (Juslin et al., 2002).5 The model com-
prised only four main sources of variability (Generative rules, Emotional
expression, Random variability, Motion principles), since we had not figured
out how to implement the fifth component, S, at the time. The model takes as
its input (a) a musical notation, and (b) a performer interpretation. We
assume that there is both a structural interpretation (e.g. phrase structure)
288 Psychology of Music 31(3)
and a mood interpretation (e.g. ‘joy’). Then, for each component, there is a
module featuring a set of algorithms that convert the notation into patterns
of expression. Space does not permit description of the rules, but some of
them are derived from the KTH rule system (see Friberg, 1995), others were
presented in Juslin et al. (2002). In a preliminary experiment, a subset of the
rules featured in the GERM model was used to synthesize expressive perform-
ances of a brief piece of music. The aim of this experiment was to examine
(1) whether the four components would yield predicted effects on listeners’
judgements of the performances, and (2) whether the components would
yield at least partly independent effects on these judgements (for further
information about this experiment, see Juslin et al., 2002). The basic idea was
to manipulate the four components of the GERM model in a factorial design;
this was done by generating all possible combinations (16) of the presence
and absence of each GERM component. For instance, a condition with only
the G-component present and all other components absent would include
only those rules that serve to convey the structure to listeners. We asked
musically trained listeners to rate all 16 performances on scales believed to
reflect various aspects of expression (clear, sad, human, gestural, musical,
expressive).
The results of this experiment suggested, first of all, that the different com-
ponents yielded predicted effects on listeners’ ratings of the performance. For
instance, the G-component yielded high ratings on the ‘clear’ scale; the E-
component yielded high ratings on the ‘sad’ scale; and the M-component
yielded high ratings on the ‘gestural’ scale. Multiple regression analyses
revealed that listeners’ judgements of overall expressivity could be successful-
ly predicted on the basis of a linear combination of the GERM components,
which all contributed significantly but in different degrees to the judged
expressivity; R = .77, F(4,184) = 66.143, p < .0001, with beta weights in
order of predictive strength: Emotional Expression, β = .64, p < .0001;
Motion Principles, β = .32, p < .0001; Generative Rules, β = .23, p < .0001;
Random Variations, β = .14, p < .01). There were, of course, some interac-
tions between the four components in how they affected listeners’ judge-
ments. However, only 20 percent of the total number of possible interactions
were actually significant, and the interaction effects were smaller than the
main effects. Finally, the components produced different and at least partly
independent effects on listeners’ judgements. This tendency is seen in exam-
ples of dissociations between the four components. For instance, the perform-
ance with only G present yielded a high mean rating on the ‘clear’ scale, but
low mean ratings on the ‘sad’, ‘human’ and ‘gestural’ scales (Juslin et al.,
2002).
into a common model. The reader may disagree with me about the specifics of
this model, but the important point is that there are different facets of expres-
sion. I suspect that researchers have found performance data confusing part-
ly because they have treated them as if they reflected a single factor. If we are
unaware of the different aspects of performance expression, or ignore some
of them, we are likely to be confused by data that reflect some unknown mix-
ture of the different modes of behaviour. Indeed, even some of the more
sophisticated systems for synthesis of music performances, such as Director
Musices (Friberg et al., 2000), may not have sufficiently differentiated between
different kinds of expressive features. The rules in Director Musices were
simply designed to make a synthesized music performance sound as good as
possible, without categorizing expressive features theoretically. This makes
perfect sense from an engineering point of view, although, unfortunately, it
muddles important distinctions between psychologically different compo-
nents of expression. Rules that reflect acoustic characteristics of certain
musical instruments are mixed with rules that convey structure (Friberg,
1995), and rules that function to convey structure are ‘turned backwards’ to
express emotions (Bresin and Friberg, 2000). A genuine understanding of
performance expression may require that we take greater care in distinguish-
ing different categories of expressive features. Distinguishing different com-
ponents of expression could help us to better explain individual differences
among performers. Different performers may be characterized in terms of
the relative weights they give to different aspects of expression. (Some may
emphasize precision, others may emphasize emotional expression at any cost,
yet others may emphasize structural communication.) Furthermore, differ-
ent musical styles may put different emphasis on different facets of expres-
sion; the G-component may be crucial to the classical pianist, but is less
important to the blues guitarist. Finally, a componential approach might help
to resolve some inconsistent findings in the previous literature. For example,
the fact that experienced listeners have shown refined ability to detect expres-
sive aspects in some studies but not others (Rohwer, 2001) could be explained
by the fact that these studies have investigated different components of
expression (e.g. generative structure vs emotional expression).
How can we be sure that the particular components I have postulated are
the appropriate ones? This is, ultimately, an empirical question, which is far
from settled. I remain open to the possibility that a different set of components
can provide a more cohesive account. However, I have argued that psycholog-
ical theory in combination with findings from listening experiments, per-
formance measurements, and brain imaging could help us to resolve this
problem. As should be apparent, I hypothesize that all components in the
GERMS model have their origins in human phenomena outside the music
domain. (From this perspective, it may be premature to view some compo-
nents as more ‘musical’ than others.) Thus, one implication of the GERMS
model is that it agrees with Palmer’s (1997) view of music performance as ‘a
290 Psychology of Music 31(3)
seemingly unique human ability that is not unique in its underlying cognitive
mechanisms’ (p. 134). But far from depreciating the performance of music,
this idea suggests that music performance is an activity that recruits many
different aspects of human behaviour in a wonderful way. It also implies that
psychological theory about various aspects of human behaviour such as lan-
guage, emotion, motor-coordination, movement and expectancy may be
highly relevant to an understanding of performance expression.
happy and sad pieces apart), we cannot ignore that most performers and lis-
teners seem to think of music as a means of emotional communication. I do
not claim that communication is the sole or main purpose of music, but 83
percent of music students in a questionnaire study claim that they try to
express specific emotions in their performance ‘always’ or ‘often’ (Lindström
et al., in press), and we know that listeners are able to recognize at least
certain emotions expressed by performers (Juslin, 2001). This suggests that
the concept of communication is viable in music, even if not every instant
of music performance may be an example of the phenomenon. To what
extent do music performers actually succeed in communicating emotions to
listeners?
In the most extensive review of emotional expression in music perform-
ance to date (Juslin and Laukka, in press), including 41 studies, a meta-
analysis of communication accuracy showed that professional performers
are able to communicate basic emotions (happiness, anger, sadness, fear, ten-
derness) to listeners with an accuracy approximately as high as in facial and
vocal expression of emotions. The overall decoding accuracy was π = 0.89,
which is equivalent to a raw accuracy score of pc = .70 in a forced-choice task
with five response alternatives (i.e. the average number of emotions included
in studies of music performance so far).6 Amateur musicians communicate
emotions less accurately, and tend to apply acoustic features inconsistently
(see Juslin and Laukka, 2000; Rohwer, 2001). The available evidence
indicates that the communicative process operates on a fairly broad level of
emotion categories, whereas finer distinctions within the categories are diffi-
cult to convey (Juslin and Lindström, 2003), at least without some context
(London, 2002; see also Juslin, 1997c). It is not that music performances
intended to express, say, anger and jealousy sound exactly the same or that
listeners cannot hear a difference, it is just that they are not able to tell which
performance is which. It has been argued by some authors that the specific
ordering of successive emotional states could help to communicate subtler or
more complex states (Levinson, 1990). This is an interesting possibility, but it
remains to be demonstrated that this is possible. (In reviewing these findings,
I do not intend to imply that performances that do not reliably convey a spe-
cific emotion are without value, because, as I have tried to demonstrate in
this article, musical expression involves a lot more than just emotion; and
sometimes musicians may deliberately aim for emotional ambiguity.)
Many studies have tried to capture the acoustic cues that musicians use to
convey specific emotions (e.g. Jansens et al., 1997; Juslin, 1997a; Juslin and
Madison, 1999; Kotlyar and Morozov, 1976; Mergl et al., 1998). These cues
involve changes in tempo, sound level, articulation, timbre, timing, tone
attack and decay, intonation, vibrato extent and frequency, accents on partic-
ular notes, etc. Earlier studies have mostly been limited to a few emotions.
Figure 2 offers examples of patterns of acoustic cues used by professional per-
formers to express 12 emotions in a recent study (Juslin and Lindström,
Juslin: Five facets of musical expression 293
Intended Emotion
Intended Emotion
Note. Sound level is expressed in decibels (dBs) down from an arbitrary reference level.
FIGURE 2 Overall levels of sound level and tempo (means and SDs) in professional pianists’
renditions of 12 emotions in performances of a theme by Haydn (adapted from Juslin and
Lindström, 2003).
294 Psychology of Music 31(3)
2003). Both ‘basic’ (e.g. sadness, anger, fear) and ‘complex’ (e.g. jealousy,
shame, pride) emotions were featured in an attempt to accommodate the sub-
tleties of musical expression. In reality the emotional expression will often
change quickly within a piece, or even within a single note. Thus we have
recently experimented with real-time analysis of emotional expression over
time (Friberg et al., 2002). That the relevant cues have been described may be
confirmed by programming synthesized performances of music on the basis
of empirical data. Computerized performances with appropriate acoustic fea-
tures can communicate emotions as reliably as human performers can
(Juslin, 1997b), although they sound less musically satisfying (presumably
because they lack other components of expression).
An important question concerns the origins of the acoustic cues used by
performers. Some authors have argued that emotions expressed in music
have nothing in common with other forms of expression: ‘Musical feelings
have their own character . . . the emotions [music] formulates are not identical
with those accompanying extra-musical experience’ (Lippman, 1953: 569).
It is getting exceedingly hard to make such a case. A review of 145 studies
(101 speech and 41 music studies) strongly indicates that patterns of musical
cues used to convey discrete emotions mainly derives from the non-verbal
aspects of emotional speech (Juslin and Laukka, in press), as proposed by
Spencer (1857). Hence, musical emotions touch us deeply not because
they are so different from everyday emotions, but because they are so similar.
This may explain why music students find extramusical sources (e.g. life
situations) useful in developing expression (e.g. Woody, 2000). One reason
why singing musical phrases seems to be such good practice for instrumen-
talists (e.g. Dubal, 1985: 221; see also Mills, 2003, this issue) could be that
this helps the performer to connect with fundamental principles of vocal
expression of emotion.7
Although it may be tempting to connect musical expression of emotion to
motion (Woody, 2002), a theoretically more plausible and parsimonious view
holds that the origin of the emotion component is to be found in involuntary
and emotion-specific physiological changes associated with emotions, which strong-
ly influence different aspects of voice production (Juslin and Laukka, 2001;
Scherer, 1986). Music performance shares with vocal expression of emotions
not just the code but also the coding: the cues used to express emotions are
uncertain but redundant to some extent, as conceptualized by the lens model
(Juslin, 2000). This can account for some interesting phenomena, for
instance that communication of emotion in music is generally successful
despite individual differences in utilization of acoustic cues among both per-
formers and listeners (Juslin, 2000) and different cues available on different
musical instruments (Juslin and Laukka, in press).
I envisage several important directions for future research. First, there has
been almost no attempt to study how the emotional expression of a perform-
ance interacts with the expression of the piece. In our recent work, we are
Juslin: Five facets of musical expression 295
using an expanded version of the lens model, which features both composed
and performed cues (Juslin and Lindström, 2003). Second, it is important
that basic research is turned into useful applications for performers. We are
thus currently trying to develop computer software that may provide per-
formers with detailed feedback about their performances (Juslin et al., in
press). Finally, it is crucial to explore how the emotion component interacts
with other components of expression in yielding an expressive performance,
which could explored by means of theoretical models (e.g. along the lines of
the GERMS model).
‘destroys the magic’, or interferes with the artistic process (Dubal, 1985:
245).
Ultimately, this could reflect the different needs of scientists and artists.
Elsewhere, I have argued for an attempt to bridge the gap between art and sci-
ence (Juslin and Zentner, 2002; see also Rink, 2003, this issue), because I
believe that musicians and researchers have much to learn from each other.
Art and science share many characteristics: they are both interesting ways of
exploring the world around us; they both involve creativity; and they both
provide certain perspectives or ways of looking at the world. However, I
believe that we must accept that there will always be some tension between
art and science, since art and science, ultimately, have different aims. Thus,
for instance, research on performance expression aims to develop general
models (i.e. simplified descriptions and explanations of a specific phenome-
non), whereas the performing artist often strives to create something unique
and personal. (The scientist usually opts out at the point where the artist
thinks things are getting interesting.) There is no way in which a psychologi-
cal focus on general models can fully capture the rich, personal and piece-
specific ways in which musicians tend to approach their work artistically.
And – after all – the most useful approach to performance as an artist may
not be the most useful approach to performance as a researcher, given the
quite different aims of these endeavours. But there is one important reason
why we should at least try to establish some common ground: the teaching of
performance expression clearly appears to benefit from explicit instruction
and knowledge about expression (Juslin et al., in press; Woody, 1999).
Hence, in this article, I have suggested that the primary goal of a psychologi-
cal approach to music performance should be to explain the nature of per-
formance expression in order provide a solid foundation for the teaching of
expressive skills in music education. It remains to be seen whether this chal-
lenge can be met in coming years. But even if it is not, I am sure that we can
learn something important along the way.
AC K N OW L E D G E M E N T S
NOTES
1. Some musicians, including Nicolo Paganini and blues singer Robert Johnson,
have been accused of ‘selling their soul to the devil’ in order to gain their excep-
tional skills.
2. One notable reason for doing so is that philosophical inquiry does not lend itself
easily to applications in music education. To take but one example: which music
teacher could seriously claim to have gained a better sense of how to teach
Juslin: Five facets of musical expression 297
expressive skills after having read Scruton’s (1997) complex chapter on musical
expression?
3. Meyer (1956) notes that ‘because Seashore advances no theory or attempts no
explanation of the relationship between deviation and affective aesthetic experi-
ence, his viewpoint lacks substance and plausibility’ (p. 203).
4. The reader may wonder why Clynes’s (1977) work has not been included here.
The reason is that independent attempts to replicate his essentic forms have yield-
ed mixed results (Gorman and Crain, 1974; Nettelbeck et al., 1989; Trussoni et
al., 1988).
5. A computational model means that the relevant relationships are expressed in
terms of mathematical procedures. The calculations are handled by implement-
ing the model in a program, which permits simulation of the phenomenon under
investigation.
6. Rosenthal and Rubin’s effect size index, pi (π), allows researchers to transform
accuracy scores involving any number of response alternatives to a standard
scale of dichotomous choice, on which .50 is the null value and 1.00 corresponds
to 100 percent correct recognition.
7. Recall also the use of principles from vocal rhetoric among composers in the 18th
century.
REFERENCES
Allison, T., Puce, A. and McCarthy, G. (2000) ‘Social Perception from Visual Cues:
Role of the STS Region’, Trends in Cognitive Sciences 4: 267–78.
Boyd, J. and George-Warren, H. (1992) Musicians in Tune: Seventy-Five Contemporary
Musicians Discuss the Creative Process. New York: Fireside.
Bregman, A.S. (1990) Auditory Scene Analysis: The Perceptual Organization of Sound.
Cambridge, MA: MIT Press.
Bresin, R. and Friberg, A. (2000) ‘Emotional Coloring of Computer-Controlled Music
Performance’, Computer Music Journal 24: 44–62.
Bryden, M. P., Ley, R. and Sugerman, J. (1982) ‘A Left-Ear Advantage for Identifying
the Emotional Quality of Tonal Sequences’, Neuropsychologia 20: 83–7.
Carreras, J. (1992) Sjunga med själen (Singing with soul). Uddevalla: Vocalis.
Clarke, E.F. (1988) ‘Generative Principles in Music Performance’, In J.A. Sloboda (ed.)
Generative Processes in Music: The Psychology of Performance, Improvisation, and
Composition, pp. 1–26. Oxford: Clarendon Press.
Clarke, E.F. (1995) ‘Expression in Performance: Generativity, Perception, and
Semiosis’, In J. Rink (ed.) The Practice of Performance: Studies in Musical
Interpretation, pp. 21–54. Cambridge: Cambridge University Press.
Clarke, E.F. and Windsor, W.L. (2000) ‘Real and Simulated Expression: A Listening
Study’, Music Perception 17: 277–313.
Clynes, M. (1977) Sentics: The Touch of Emotions. New York: Doubleday.
Cohen, R.A., Riccio, C.A. and Flannery, A.M. (1994) ‘Expressive Aprosodia
Following Stroke to the Right Basal Ganglia: A Case Report’, Neuropsychology 8:
242–5.
Cook, N. and Dibben, N. (2001) ‘Musicological Approaches to Emotion’, in P.N. Juslin
and J.A. Sloboda (eds) Music and Emotion: Theory and Research, pp. 45–70. New
York: Oxford University Press.
Davidson, J.W. and Correia, J.S. (2002) ‘Body Movement’, in R. Parncuttand G.E.
McPherson (eds) The Science and Psychology of Music Performance: Creative
298 Psychology of Music 31(3)
Strategies for Teaching and Learning, pp. 237–50. New York: Oxford University
Press.
Davies, S. (1994) Musical Meaning and Expression. Ithaca, NY: Cornell University
Press.
Dubal, D. (1985) The World of the Concert Pianist: Conversations with 35 Internationally
Celebrated Pianists. London: Victor Gollancz.
Eibl-Eibesfeldt, I. (1989) Human Ethology. New York: Aldine.
Epstein, D. (1995) ‘A Curious Moment in Schumann’s Fourth Symphony: Structure
as the Fusion of Affect and Intuition’, in J. Rink (ed.) The Practice of Performance,
pp. 126–49. Cambridge: Cambridge University Press.
Farga, F. (1969) Violins and Violinists, 2nd edn. London: The Cresset Press.
Friberg, A. (1995) ‘A Quantitative Rule System for Musical Performance’, doctoral dis-
sertation, Royal Institute of Technology, Stockholm.
Friberg, A. and Sundberg, J. (1999) ‘Does Music Performance Allude to Locomotion?
A Model of Final Ritardandi Derived from Measurements of Stopping Runners’,
Journal of the Acoustical Society of America 105: 1469–84.
Friberg, A., Colombo, V., Frydén, L. and Sundberg, J. (2000) ‘Generating Musical
Performances with Director Musices’, Computer Music Journal 24: 23–9.
Friberg, A., Schoonderwaldt, E., Juslin, P.N. and Bresin, R. (2002) ‘Automatic Real-
Time Extraction of Musical Expression’, in Proceedings of the International Computer
Music Conference, Göteborg, 16–21 September 2002, pp. 365–7. San Francisco:
International Computer Music Association.
Gabrielsson, A. (1987) ‘Once Again: The Theme from Mozart’s Piano Sonata in A
Major: A Comparison of Five Performances’, in A. Gabrielsson (ed.) Action and
Perception in Rhythm and Music, publication no. 55, pp. 81–103. Stockholm: Royal
Swedish Academy of Music.
Gabrielsson, A. (1999) ‘The Performance of Music’, in D. Deutsch (ed.) The Psychology
of Music, 2nd edn, pp. 501–602. San Diego: Academic Press.
Gabrielsson, A. (2003) ‘Music Performance Research at the Millennium’, Psychology
of Music 31(3): 221–72.
Gabrielsson, A. and Juslin, P.N. (1996) ‘Emotional Expression in Music Performance:
Between the Performer’s Intention and the Listener’s Experience’, Psychology of
Music 24: 68–91.
Gabrielsson, A. and Juslin, P.N. (2003) ‘Emotional Expression in Music’, in
R.J. Davidson, H.H. Goldsmith and K.R. Scherer (eds) Handbook of Affective Sciences,
pp. 503–34. New York: Oxford University Press.
Gilden, D.L., Thornton, T. and Mallon, M.W. (1995) ‘1/f Noise in Human Cognition’,
Science 267: 1837–9.
Gorman, B.S. and Crain, W.C. (1974) ‘Decoding of “sentograms”’, Perceptual and
Motor Skills 39: 784–6.
Hallam, S. (1995) ‘Professional Musicians’ Approaches to the Learning and
Interpretation of Music’, Psychology of Music 23: 111–28.
Ivry, R.I., Keele, S.W. and Diener, H.C. (1988) ‘Dissociation of the Lateral and Medial
Cerebellum in Movement Timing and Movement Execution’, Experimental Brain
Research 73: 167–80.
Jansens, S., Bloothooft, G. and De Krom, G. (1997) ‘Perception and Acoustics of
Emotions in Singing’, in Proceedings of the Fifth European Conference on Speech
Communication and Technology, Vol. IV, pp. 2155–8. Rhodes, Greece: ESCA.
Johansson, G. (1973) ‘Visual Perception of Biological Motion and a Model for its
Analysis’, Perception and Psychophysics 14: 201–11.
Juslin: Five facets of musical expression 299